Monitoring and Estimating Species Richness

1. Monitoring and Estimating Species Richness Paul F. Doherty, Jr. Fishery and Wildlife Biology Department Colorado State University Fort Collins, CO

2. Background • Species diversity or richness may be an important objective of monitoring/management programs • Especially with careful definitions • e.g. maximizing the number neotropical migrants, bats, frogs

3. Background • In estimating species diversity the most common measures are: • Measure of evenness (relative abundances) • Simpson’s Index • Shannon Index (a.k.a. Shannon-Weiner, Shannon-Weaver) • Species richness (presence/absence) • No need to tell individuals apart • Often-made assumptions • detection rates = 1 • all species sampled with the same probability. • In estimating species diversity the most common measures are: • Measure of evenness (relative abundances) • Simpson’s Index • Shannon Index (a.k.a. Shannon-Weiner, Shannon-Weaver) • Species richness (presence/absence) • No need to tell individuals apart • Often-made assumptions • detection rates = 1 • all species sampled with the same probability.

4. Challenges • Different species will have different detectabilities • Certainly some species will be missed • e.g. Compare a thrush with a gnatcatcher song • Abundance will vary for each species • To use Shannon or Simpson’s Indices • must estimate abundance of each species • Not clear how to interpret these indices (notions of evenness and dominance) • Focus on specie richness • Easier to tell species apart than individuals • In past studies species richness and these indices are often highly correlated.

5. Direct connection in estimating richness to estimating abundance • Instead of estimating abundance (# of individuals), estimate species richness (# of species) • Instead of estimating survival, estimate extinction, etc.

6. Time scale • Local or global • Local probably most relevant in our case

7. Estimation • Importance of detection probability (C = count, p = probabilty of detection, S = richness) • Relative change – issue does not go away

8. Estimation • Can use ‘heterogeneity’ estimators such as Burnham and Overton’s (1978,1979) Jackknife. • Makes sense to use estimator such as this. • Desirable to have equal sampling effort (e.g. quadrats, time spent listening). • Can be used with ‘capture history’ or X matrix • N = number of species detected, t = number of quadrats sampled, f = number of species found on only one quadrat • For stats-minded folks – this is probably one area where growth will occur

9. Sampling • Sampling over time or space • Capture history for each species • 010110 • Can generate frequencies (# of species seen exactly once, twice, etc) • Can estimate richness using widely available software (Capture, SpecRich, Comdyn). T1 T2 T3 T4 T5 T6 L2 L1 L3 L4 L6 L5

10. closed closed closed open open Community dynamics • Change in species richness • Extinction • Turnover • Colonization • Use Pollock’s Robust Design X1X2X3 Y1Y2Y3 Z1Z2Z3

11. Community dynamics • Change in species richness over time • Not changes in composition, just numbers • ‘Stability’ (temporal variance)

12. Local Extinction • Probability that a species present in sampling period i is not present in a later period j.

13. Local Species Turnover • Probability that a species selected at random from the community at time j is a “new” species as compared to the community at time i. • 1 if every species is “new” • 0 if every species is “old”

14. Number of colonizing species • The number of species not present at time i that colonize and are present at time j.

15. Spatial Analogs • Comparing two sites • Relative species richness • Species co-occurance and omission • Comparing reference and altered sites • (Forest Plan)

16. Example • Comdyn, SpecRich • http://www.mbr-pwrc.usgs.gov/software.html#comdyn

17. Example • Compare a BBS route (WI in 1970 and 1990)

18. Summary • Community richness and dynamics can be estimated nicely and easily • May be useful in monitoring programs • Especially where monitoring concerns are ‘broad’. • Data are often easier/cheaper to collect than those associated with abundance

19. Example:Sexual Selection affects local extinction and turnover in bird communitiesDoherty P.F., Sorci, G., Royle, A., Nichols, J.D., and T. Boulinier

20. Background • Sexual selection is thought to drive the evolution of secondary sexual traits such as bright colors and plumage dimorphism

21. Benefit of these traits • Increased reproductive success (Andersson 1994)

22. Costs of such traits • Increased predation and sensitivity to environmental and demographic stochasticity (Haskell 1996, Legendre et al. 1999) • Prediction: Extinction rates should be higher for species with intense natural selection (Tanaka 1996)

23. Support for prediction • Dichromatic bird species have been shown to have higher mortality rates than monochromatic (Promislow et al. 1994) • Dichromatic bird species introduced onto islands went extinct at a higher rate than monochromatic species (McLain et al. 1995, Sorci et al. 1998)

24. Methodological Issues • Detection error • Dimorphic and monomorphic species may not be detected with equal probability • Differ with habitat • Temporal differences • Variation in space • Local populations close together may be synchronous • Spatial autocorrelation shown to be important is island study of species richness (Selmi and Boulinier 2001)

25. Objective of our study • To test the prediction that dichromatic species have higher local extinction rates • Improvements • Also examined turnover rates • Tested on continental scale. • Incorporated detection probabilities • Incorporated spatial autocorrelation

26. Data set • US Breeding Bird Survey data (1975-96) • Restricted dataset to daytime species from 6 orders and grouped them as either having dichromatic (n = 153) or monochromatic (n=185) plumage.

27. Definition • Extinction - proportion of species becoming locally extinct (on a BBS survey route) between two successive years among species present the first year. • Turnover – proportion of new species among species present the second year as compared to the first

28. Detection Error • We used COMDYN (1999) to estimate species richness, extinction rate, and turnover rate for dimorphic and monomorphic species • This program and associated metrics was developed to better estimate these quantities as well as detection probabilities • Based on ‘closed’ capture models with heterogeneity (Mh) and robust design (Boulinier et al. 1998, Nichols et al. 1998)

29. Spatial modeling • We used a Bayesian approach • Computed the difference between dichromatic and monchromatic extinction rates

30. Spatial modeling • T-test would assume yi independent normally distributed • However, yi‘s probably not independent due to spatial proximity • We could resort to a subjective stratification scheme with associated restrictive forms of spatial dependence • Or we could introduce spatial correlation into the problem explicitly by adding an error term which we assume is spatially correlated

31. Normally distributed random variable Correlated spatial random variables Spatial modeling • Many ways to parameterize the spatial correlation among the αi’s • Could insert parameters from a variogram (parametric approach) • We used a convolution approach suggested by Higdon (1998)

32. Results • Dichromatic species had a 23% higher local extinction rate per survey than monochromatic species (0.079 ± 0.001 vs. 0.064 ± 0.001) • Widespread result across continent indicating that sexual selection influences the communities of birds in many differing habitats and places. • We also found that a model incorporating a spatial component fit better than one without (ΔAIC = 117), and that we should qualify above statement

34. Mean difference in extinction rate Lower 95% CI

35. Results • If extinction rates are higher in dichromatic species, then some might expect the number of dichromatic species to decrease over time. • We found the number of species in both groups increased over the time period (4.8% and 3.4% respectively). • We examined turnover rates too

36. Mean difference in turnover rate Lower 95% CI

38. Conclusions • Higher local extinction and turnover rates for dichromatic vs. monochromatic species • Spatial structure in these rates suggest metacommunities (Wilson 1992) • Regionally connected communities • Regional or local factors may have dramatic effects over a large area.